Dedicated GPUs for production model hosting
Rent whole GPUs from a prepaid wallet, pick any GGUF model, and get an OpenAI-compatible endpoint on hardware that is yours alone for the whole term.
Specifications
| GPU | NVIDIA RTX PRO 6000 Blackwell |
|---|---|
| Memory | 96 GB GDDR7 per GPU |
| API | OpenAI-compatible |
| Billing | Prepaid, hourly or monthly |
| Availability |
7 of 7 GPUs available
|
Pricing
Simple pricing per dedicated GPU. Hourly for experiments, monthly for production. Prices in BDT, charged from your wallet.
| Plan | Starter 1×GPU | Pro 2×GPU | Scale 4×GPU |
|---|---|---|---|
| GPUs | 1 | 2 | 4 |
| Total GPU memory | 96 GB | 192 GB | 384 GB |
| Hourly price | BDT 250.00 / hour | BDT 480.00 / hour | BDT 940.00 / hour |
| Monthly price | BDT 120,000.00 / month | BDT 230,000.00 / month | BDT 450,000.00 / month |
| Action | Deploy this plan | Deploy this plan | Deploy this plan |
GPU inventory
Availability updates as deployments start and stop. Deployment names are not shown.
Dedicated hardware
Each deployment gets whole GPUs for the full term. Nothing else is scheduled onto them.
OpenAI-compatible endpoint
Every deployment gets its own URL and API key. Point existing OpenAI client code at it.
Transparent prepaid billing
Top up once, then pay per hour or per month from the wallet. Every charge has an invoice.
How it works
From top-up to a live endpoint in three steps.
-
Top up your wallet
Send a bKash, Nagad or bank transfer and submit the reference. An admin verifies it and credits your balance.
-
Choose GPUs and a model
Select 1 to 4 GPUs, an hourly or monthly term, and a model from the shared library or Hugging Face.
-
Call your endpoint
Your deployment goes live with its own key and URL. Use it like any OpenAI-compatible API.
Frequently asked questions
Is the GPU shared with other customers?
No. Each deployment gets dedicated GPUs for the whole term. Nothing else is scheduled onto them.
How do I pay?
Top up your wallet by bKash, Nagad or bank transfer, then submit the transaction ID. Deployments are charged from the wallet balance.
Can I use my own GGUF model?
Yes. Pick a model from the shared library or point to any public Hugging Face GGUF file. Models are served with llama.cpp.
What happens when my term ends?
The deployment is stopped and its GPUs are released. Renew before expiry to keep the same endpoint and key.